Events always take (place with) ser

2.1 Object nouns and event nouns

Following the classification proposed by Fábregas and Marín (2012), we call ‘Object Nouns’ those nouns denoting any physical entity: either concrete objects, such as table or house, or animate individuals, such as teacher or Rosalía (9). On the other hand, ‘Event Nouns’ refer to non-physical entities that have a spatial and temporal anchor; i.e., entities that ‘take place’ (10).

Note that in (9) and (10a) we have focused on underived nouns; some deverbal nominalizations (10b), of course, also denote events (Grimshaw’s 1990 ‘complex event nouns’), which pattern like the nouns in (10a) in terms of locative copulas; other nominalizations (Grimshaw’s 1990 ‘result nouns’) denote participants in the event (11), and as such, they pattern with (9):

As explained before, Event Nouns need to be located with ser (12a) and Object Nouns with estar (12b); this copula specialization holds strong for definite nous, but when locating an indefinite noun, -so no definiteness effects arise-, being the indefinite an event or an object, then both types of nouns combine with the same existential predicate, haber. Thus, the specialization is idiosyncratic to copular verbs:

As already advanced, a subset of nouns such as cena ‘dinner’, examen ‘exam’, fuegos artificiales ‘fireworks’, obra de teatro ‘theater play’ can be located with both copulas, ser and estar, depending on their interpretation (Perpiñán et al. 2020). When clase ‘classroom’ refers to the physical space, it needs to combine with estar (14b), but when it refers to the lecture given during that class it combines with ser (14a).

In the following paragraphs, we will present the most relevant differences and similarities between these two types of ontological entities in order to later account for their distribution with the Spanish copulas.

Objects and events share the property of being countable: dos mesas ‘two tables’; tres reuniones ‘three meetings’, and they both can be located, but they differ in several other properties. As shown in the following examples, taken from Huyghe and Azzopardi (2019), objects can be modified in their different physical dimensions, as in (15), whereas events, as abstract entities, cannot, as in (16).

Objects and events also differ in their temporal dimension. Observe, first, that events (17) convey a temporal extension that objects lack (18).

Second, event nouns – but not object nouns – can function as temporal frames or locators and, hence, be introduced by temporal prepositions, as in the following examples, also taken from Huyghe and Azzopardi (2019).

Third, event nouns can combine with predicates such as verse interrumpido ‘to be interrupted’ or modifiers such as en curso ‘in progress’ (Fábregas et al. 2012); a possibility that objects do not exhibit. Verse interrumpido requires an internal argument that exhibits some temporal progression so that it can be interrupted before its natural end, and this is only satisfied by event-denoting nouns that involve a dynamic change. In the same way, en curso requires internal dynamic change that physical entities lack.

But probably the most significant difference between these two types of nouns is that events take place, both in space and time, (23), while objects, (24), do not.

From here – the fact that events but not objects can be located in time – it naturally follows that ser is the only copula that can locate in time; by the same token, estar is not compatible with expressions that locate in time.

2.2 State nouns and quality nouns

According to several authors (Arche and Marín 2014; Fábregas 2016; Jaque and Martín 2019), in addition to individuals and events, there are at least two other ontological noun types: states (26) and qualities (27). The nouns in (26) are used to denote temporal periods associated with eventualities, but lack any type of dynamicity or change: either when they denote the resulting state or other types of states, the nouns in (26) are not events.

With respect to (27), they denote the abstract name of a quality, defined typically by their morphological base, which is an adjective.

Notably, these two types of nouns have an uncontroversial abstract denotation at least in the sense that they do not denote physical objects with a spatial location, a characteristic that has sometimes been adduced to explain the different behavior of events in comparison to objects and in particular to explain its late acquisition in L1 and L2 (Sera 1992). From this perspective, one could expect that estar will not be the copula to locate them, especially if estar selects physical objects as the subjects it locates. Thus, if estar is the only copula with selection restrictions and ser is the default, it is surprising that both states (28) and qualities (29) always require estar in locative sentences, despite its abstractness.

Considered from this perspective, then, the question is what makes events incompatible with estar. With this in mind, we will now describe the properties of state and quality nouns compared to those of object and event nouns. Only with a complete picture of the grammatical characteristics of all ontological types of nouns that can be located, will we be able to properly account for the place events occupy in this classification. We will test state and quality nominals with the same diagnostics used to describe objects and events in the preceding section.

Observe, first, that both states (30a), and qualities (30b), are not count nouns, unlike event nouns, particularly those that are underived from verbs (31).

States and qualities, unlike events, take place neither in space nor in time:

Event nouns can combine with predicates such as verse interrumpido ‘to be interrupted’ or modifiers such as en curso ‘in progress’ (Fábregas et al. 2012); this is not possible with states or qualities.

As we have seen in the preceding section, event nouns are compatible with temporal locators; in contrast, neither states nor qualities are, as they do not denote eventualities that can progress along a temporal scale:

However, this does not mean that quality and state nouns do not contain any type of temporal information. While it is still a matter of debate whether quality nouns convey some type of temporal information (cf. Zato 2020 and references therein), there is little doubt about the temporal dimension of states, as shown in (39), where they allow temporal modifiers introduced as prepositional phrases. Some partial evidence for the temporal information of quality nouns might come from the fact that, like state nouns and event nouns, they can be the subject of a verb like durar ‘to last’.

The fact that state nouns and possibly quality nouns, like event nouns, have a temporal dimension is very informative regarding whether a subject is located with ser or estar. It is particularly illuminating to compare event nouns and state nouns, as they both belong to the set of eventualities (Bach 1986). Attending to the fact that events are located with ser and states are located with estar we have to conclude that it is not the temporal dimension per se that triggers events to combine with ser. As we have seen, state nouns, despite having a duration and allowing temporal anchoring, combine with estar. In this respect, the main difference between these two classes of nouns is that events include in their temporal denotation the idea of change, action or progression, which is completely absent from the case of state nouns. Therefore, the main difference between these two classes of nouns is their inner aspect, not their coarser properties, such as whether they are concrete or abstract or even whether they are bounded or not (Table 1).

In a nutshell, events constitute an altogether different class of nouns, presumably because only eventive nouns include in their denotation the idea of internal phases in their development, as emphasized – among others – in Grimshaw (1990) or Borer (2013). That is, event nouns are dynamic in a strict sense.

As a conclusion of this section, the behavior of state nouns and quality nouns shows that all nouns except for events are located with estar. This contradicts the general view that ser is a default copula that lacks any selectional capacity: in locative contexts, ser positively selects events and estar combines with anything else. One is then tempted to propose the generalization in (41).

Note in this respect that the location of an event is eventive itself. Assuming that the progressive periphrasis -always with estar- is a good test for eventivity, note that (42) behaves as an eventive location, while the location of anything that is not an event is stative. We hasten to nuance the claim about (42) being an eventive location to emphasize that, of course, (42) does not involve an eventive location in the sense that there is a displacement across a trajectory (in fact, see Section 3.1 below about the so-called path locations); however, it involves an eventive location in the sense that there is a progressive change or action that is identified by its temporal and spatial location.

The claim in (41) is theoretically undesirable. It would imply that we have one default copula in the locative domain (estar) and a second default copula in non-locative uses (ser). This would make it impossible to unify the uses of copulas because it would force us to treat locative copular sentences as different from other types of copular predicates. Still, we want to preserve the idea that ser is always the default copula, but at the same time establish as an empirical fact that there is a real correlation between ser and events that requires an explanation.

At this point, two questions emerge. The first one is whether it is possible to identify other instances of ser with events beyond the purely locative context that has just been revised, and the second is how the association between events and ser can be made compatible with the general observation that ser is an IL copula and, more generally, a default copula in any other type of copulative structure. The next section is devoted to the first question, which we believe should be answered affirmatively: there are other contexts where ser combines with events.

3.1 Path copular sentences

Thus far, we have drawn the following generalization: objects, as well as states and qualities, are located with estar, while the location of events requires ser. However, this is not completely accurate, as certain objects may combine with ser when their interpretation is deictically linked. The following sentences, taken from Brucart (2012), exemplify the phenomenon we refer to:

According to Brucart (2012: 33), this pattern involves a path interpretation of the relevant preposition where the entity in the subject position, although not an event, is located at the end of the path that is defined in the post-copular expression. For Brucart (2012) this path interpretation is what makes the use of ser available. Although quite productive, in Brucart’s description, the pattern is constrained by pragmatic conditions: (i) “it usually implies a prior request for information about the location of the corresponding place by the hearer”; (ii) “the subject typically denotes a fixed place where certain activities and services take place”; (iii) the utterance situation should be “deictically anchored to the entity that is spatially located”, and perhaps most crucially, (iv) “the sentence denotes the route that the hearer must follow to reach the goal of the path” (Brucart 2012: 33). For us, the crucial property is the last one: the hearer must follow the path to reach the location as a goal. The question is whether this dynamic interpretation is part of some kind of pragmatic condition, as Brucart proposes, or not. If there is such a pragmatic requirement, the sentences in (45) take ser because of the presence of a path; if there is not such a pragmatic requirement, then the presence of an event reading – even if the event is intended and not instantiated – determines the presence of ser. We believe that the second explanation is the correct one. Importantly, paths can be combined with both copulas, as shown in (46) (see Zagona 2012).

In both cases, the path preposition is interpreted in a way similar to what Svenonius (2010) calls ‘end-of-journey’ or ‘Cresswellian locations’, that is, the figure is located at the end of the path that the prepositional structure defines. Zagona (2012) proposes that in sentences like (46) the preposition lacks the path component, but it is clear that this path component is interpreted and the figure is located at the end of that intended path that extends to a location close to the bridge. Thus, the notion of path itself cannot be what distinguishes (46a) from (46b). And still, the interpretation of (46a) is different from (46b), where we understand that an unidentified speech participant is supposed to follow that path. Our claim is that what makes (46a) and (46b) different is that only in (46b) does one interpret that there is an intended event that is required to reach the intended location: in (46b), but not necessarily in (46a), the addressee not only must have asked for directions to arrive to the pharmacy, as Brucart already noticed, but she also had the intention of going to that place, presumably to do something there – a secondary point that we will get back to in a moment.

In fact, it is possible to license the alleged path copular structures with ser in contexts where there is absolutely no path at all to follow, but where the idea of the presence of an intended event is valid. Consider (47).

(47) can be an exchange where someone is looking for a pharmacy and asks somebody at the door of a shop; that person answers that the addressee is already in the location of the pharmacy. That is: there is no further path that has to be traversed to arrive at the pharmacy, but there is an intended arrival, which seems to license the ser copula in this context. Secondarily, the notion of intended event is present in these constructions, which we now know are wrongly called ‘path copular sentences, through the nature of the subject. These types of sentences, as pointed out by Salazar (2002), tend to sound natural to the extent that the figure located is in a place where one typically performs some kind of action or event, which can be ‘buying’ (as in 47), ‘visiting’ (48a), ‘receiving a treatment’ (48b) or many other types of actions. Locations that are difficult to conceptualize as places where one can perform an event are more difficult to accept unless one contextually assigns an event function to them (as in [49c], if we decide that rock is the goal in a marathon or determine that it is a touristic attraction that one visits).

This – that the copula ser is only licensed if the interpretation implies an intended event, be it a path that has to be traversed or some action that one performs in the location if the path disappears – is how we interpret the following Brucart’s observation: “if ser is commuted by estar in these sentences, the notion of route is weakened and the new construction becomes an ordinary locative attributive” (Brucart 2012: 33). Hence, sentences of the type of (50), including a directional preposition, can only accept ser:

In our view, the reason for this contrast is that the path preposition forces the interpretation of an intended event that involves the interlocutor, at a minimum ‘arriving at the airport’. This forces an event interpretation, and like in the case of the locative copular sentences, the copula used must be ser. Thus, if sentences with ser like (45) can be licensed by a path interpretation, this is because the notion of path can be used to give content to the intended event that the copula requires, not because the path itself licenses that copula. So-called ‘path copular structures’, we believe, should be rather called ‘intended event copular sentences’ (see also Sánchez Alonso 2018, who independently reaches a similar conclusion based on a different argumentation; see also Section 4.2 below), even if the path reading is the easiest way to give content to the intended event. Likewise, we argue, there might be a second set of structures that, like locative copular sentences, involve an event interpretation of the whole clause, only that in this case it would be an intended event.

3.2 Passives and structures with participles

According to Coussé (2011) and Marín (2016), among others, participles of telic verbs – i.e., those verbs including an endpoint (a telos) – can denote both events and states; it depends on the temporal perspective adopted with respect to the boundary. In certain cases, as in (51), the temporal perspective ends at that point, since it is denoted by an event that culminates. In other cases, (52), the perspective adopted starts just after that point, since it is denoted a resultant state.

As illustrated, eventive passives are built with ser (51), while stative passives are built with estar (52) (Bosque 1990; Fernández Leborans 1995; Jaeggli 1986; Luján 1981, among many others). This is what we observe for telic predicates, whose participles have the double possibility to denote the resulting state of a previous event, as in (52), or to denote the complete event, as in (51). Crucially, in the case of activities or processes, only eventive passives are available because the participle does not have the possibility to denote the resulting state that follows the completion of the event, as these events are atelic and lack a natural endpoint. Thus, only ser is allowed because the participle must denote an event.

In the passive, then, we have another context where ser directly associates to event interpretations, this time not expressed by the subject but rather by the complement of the copula. As in the case of locative copular sentences, this property should be a priori unexpected if the only information that the copulas care about is the distinction between IL and SL predicates. This would be incompatible with a view in which ser is the default member of the set of copulas and lacks any type of selectional restrictions.

Of course, we should keep in mind that, in general, Individual Level predicates are associated with the copula ser, and Individual Level predicates are the purest form of non-eventive element, something that is – in principle – difficult to integrate with the present empirical overview, where ser is related directly to eventive interpretations. The picture is, then, extremely puzzling: what should IL predicates have in common with events, to the exclusion of SL predicates?

Then, we move on to the second problem mentioned at the end of Section 2, namely, whether it is possible to integrate this empirical picture with the more general uses of the copulative verbs in other predicative contexts. In order to do so, we will revise a number of approaches to the locative copula and the nature of copulative verbs in general, to see which ingredients they provide as tools to potentially explain this connection.

4.1 Unifying approaches based on underspecified ser (1): prepositional nature

There is a first family of analyses that try to account for locative copular sentences by integrating them with other uses of the copula within a system that adopts one single aspectual distinction that is based in the IL/SL or the imperfective/perfective dichotomy. Two works that, from different perspectives, fall into this class are Brucart (2010, 2012 and Zagona (2012).

The two works share the crucial assumption that the copula ser is the underspecified copula, which emerges when some additional information is missing or when that additional information has been licensed by some other element, typically an internal property of the predicate or subject.

Moreover, their analyses relate estar to some prepositional content. In both Brucart and Zagona, but with different perspectives, estar is marked by a prepositional feature that needs to be licensed by material in the copula, following initial intuitions of Benveniste (1966) that treat some more complex light verbs as manifestations of a default copulative verb combined with a preposition.

Brucart (2012) proposes to reinterpret the traditional IL/SL dichotomy in a locative framework by assimilating it to the central coincidence versus terminal coincidence contrast in the prepositional domain (Hale 1986; Hale and Keyser 2002). This way, individual-level predicates would be characterized by central coincidence relations, and stage-level predicates would correspond to terminal coincidence relations, which are also the ones used to express paths of motion. Following Gallego and Uriagereka (2016), Brucart proposes that estar is the copula that licenses a terminal coincidence relation, and is therefore introduced when the complement of the copula contains such preposition (54). Ser is unmarked with respect to the coincidence relation and therefore emerges whenever the terminal coincidence relation is not present or has been licensed by another element, which is the crucial property in the context of locative copular structures.

The configuration illustrated in (54a) corresponds to the combination of estar with a stage-level adjective. The structure includes a terminal coincidence preposition contained in the complement of estar, where the assumption is that a predication structure marked as terminal coincidence is projected above the predicate. This terminal coincidence predication is licensed by the copula estar, which contains an ad hoc feature to license this type of relation. Brucart, furthermore, proposes that all locations contain a terminal coincidence preposition. When the located entity is a physical object, the estar copula has to be introduced in order to license the terminal coincidence relation.

The use of ser with event subjects and with path subjects emerges because in those configurations, even though the locative structure introduces a terminal coincidence relation, the subject that is located in the prepositional structure is able to license the terminal coincidence relation before the copula is introduced (2012: 18).

For (56b), Brucart’s proposal is that the event noun in the subject position is initially introduced as the specifier of the terminal coincidence relation. This subject, denoting an event, contains a feature R_T that licenses the locative relation. This makes it unnecessary to introduce estar to license this feature. Consequently, ser is introduced as a default copula. Similarly, in (56c), the subject is interpreted as a path and therefore also contains a feature R_T, which again licenses the locative relation without estar. In this spatial theory of aspect, then, terminal coincidence relations identify with paths of movement, and events are defined as containing a path that is mapped as the progression of the event itself.

Brucart’s account has several shortcomings. We have already argued in Section 3.1 that path copular sentences are in fact possible without any type of relevant path interpretation. One could also raise questions about the syntactic locality of the relation that allows a DP subject to license the R_T feature of the prepositional structure: presumably, event nouns, which are projections headed by determiners and quantifiers, would carry the information about their eventive nature in the lexical N layer, that is, this information would be embedded within a complex constituent in a specifier position, which is a configuration that has been treated as a syntactic island in many approaches (cf. Uriagereka 1992, for instance).

However, perhaps the main problem of Brucart’s analysis is to know the nature of this R_T feature, and how its content maps onto identifiable and recognizable properties of the structures involved. It is crucial in Brucart’s account that all kinds of locative relations are defined by the feature R_T even though they seem to be basically stative, while at the same time, the feature R_T must be contained in event nouns but not in other nominals with temporal properties, such as state nouns. Remember from Section 2 that state nouns are also located with estar, which means, for Brucart’s approach, that they cannot contain the feature R_T. It is less than optimal that R_T is a property that needs to identify all locative structures, but at the same time, it is severely restricted in the nominal realm to only event nouns and nouns interpreted as paths.

Zagona’s (2012) approach shares with Brucart’s the intuition that the copula ser is a default element, and in her approach ser inherits or projects whatever aspectual information is contained in its complement. The analysis proposes that the range of complements that estar can select is determined by a formal feature [uP] estar has, and forces it to combine with a prepositional structure.

Ser would then be introduced as a default option. Zagona (2012: 305) assumes a relational theory of aspect and tense (Klein 1994) where, categorially and configurationally, aspectual relations are expressed through prepositional structures, and proposes that estar has a unified selectional capacity, selecting an abstract aspectual P – for instance, when used as an auxiliary in the progressive (58) – or a locative P, which explains the extended use of estar in locative contexts even when the interpretation is not an SL one.

In terms of the nature of [uP], Zagona (2012) states that estar only selects for (stative) prepositions expressing a single location. Complex prepositions that include a path or a direction are not compatible with estar. This is why, Zagona argues, (59a) is grammatical, while (59b) is not: in her approach, a is a path preposition equivalent to ‘to’, and estar forces the path prepositions to be absent from the structure.

In (59b), the [uP] feature of estar cannot be checked by a. According to Zagona (2012), this also explains why estar is ungrammatical with PPs that locate events (60): like in Brucart’s analysis, the eventive nouns contain a feature equivalent to a path preposition, and this feature is incompatible with licensing the [uP] feature of estar.

Along the same lines, Zagona (2012) argues that the participial phrases in (61a) and (61b) differ in structure in a manner analogous to the location versus directional PP sentences discussed above, with eventive readings having a path preposition and stative readings having a place preposition and no path preposition. Since eventive passives contain a path, only ser can be chosen as an auxiliary verb.

Zagona’s analysis faces a crucial problem, similar to Brucart’s, with respect to the configurational relation where the path preposition interferes with the presence of estar. In fact, for Zagona (2012) it should be crucial that path prepositions are ungrammatical with estar complements, but we already saw (Section 3.1) that this is far from being evident, and in fact prepositions that normally denote paths give a Cresswellian End-of-journey reading with estar. This is in principle problematic for her analysis, as she would need to derive the semantic interpretation involving a location at the end of a path from a structure that in theory should not contain a path structure. Of course, Zagona (2012) could argue with Svenonius (2010) that the Cresswellian location reading is caused when the path preposition is embedded under a locative preposition (62a), and that in such context the path preposition does not block the presence of estar because it is embedded under a more complex constituent. However, if (62a) saves the derivation, it is unclear why the presence of a (temporal) path in event nouns would be incompatible with estar, as presumably that feature, contained in the lexical layer of the nominal constituent, would be embedded under determiners, quantifiers, and potentially significant chunks of functional structure (62b).

Finally, a problem of Zagona’s analysis that Brucart’s does not have relates to the default presence of ser. Note that in Brucart (2012), ser is the default because it does not license any other structure; if the structure that the copula combines with needs to be licensed, estar is introduced. In contrast, in Zagona (2012), ser is the default because it does not need to be licensed by another structure; the structure below the copula does not need to be licensed by estar in any case. In practice, this means that in Zagona’s analysis ser could combine with locative structures in any case, and thus that there is no direct way to block (63).

Zagona acknowledges this problem and offers two preliminary suggestions, one of which is that the structure of the locative copula might be different from the one used otherwise, defining estar as the main locative copula. The second possibility is that ser, being the default, can be introduced when no more specific element (in this case, estar) is introduced, but note that in her analysis nothing of the PP structure calls for the presence of estar – it is the opposite, estar calls for the prepositional structure.

4.2 Unifying analyses based on underspecified ser (2): anchoring to an external situation or contextual parameter

There are also analyses that argue that ser lacks a specification that estar contains, but do not play around with prepositional contents. Here we will highlight two works, one that proposes that estar spatiotemporally anchors the predication to a reference situation (Escandell 2018), and one that uses the notion of contextual boundedness (Sánchez Alonso 2018).

The starting point of Escandell (2018) is that, within the domain of adjectival predicates, estar introduces the presupposition that there should be a relevant and specific situation that the predication relates with; in contrast, the function of ser is only to support the predicative relation. Adopting the notation in Maienborn (2005), Escandell (2018; see also Arche 2006; Escandell and Leonetti 2002) adopts the entry in (64b) for estar, where, in contrast to ser, estar introduces an abstract relation A between the predication e and an external situation s _e .

When estar combines with a stage-level predicate, which in itself predicates from spatiotemporal stages or slices of the individual, there is a complete matching in features. When, on the other hand, the predicate is the individual level, there is a mismatch in features that is resolved by the meaning of estar imposing a coerced interpretation of the predicate forcing it to adopt a stage-level-like interpretation. The lexical properties of the adjective do not get altered, and the combination between estar and the adjective is a case of semantic composition where the copula imposes its interpretation. This makes it possible to explain so-called evidential uses of estar (Mangialavori 2013), where estar does not impose the reading that the properties are temporally bound when exhibited by the subject, or that there is any comparison between stages of the individual, but rather that the speaker commits herself to the belief that the property is truthfully predicated from the subject and that she has first-hand experience of that property (Escandell 2018: 79). In (65), the speaker signals that she has tasted the paella and has concluded, in this first-hand experience, that is exquisite.

It is difficult, from our perspective, to understand exactly how this account can cover the location of events and the path of copular sentences that have been described above. The use of estar in the location of individuals, states, properties and so on might be included in the analysis if, along Brucart (2012), one proposes that any location act is in fact an act that requires to take as a point of reference an external entity, defining as ‘external’ the reference point employed to locate the subject. However, the question that would emerge is what makes that reference point different when the located entity is an event: would that reference point, then, be internal to the situation and not external? The approach might be able to make a claim along these lines if it accepts the core claims of Leonetti (1994) (see Section 4.5 below), where the temporal and locative dimensions are inherent properties of events that are not predicated from them. If it is possible to extend the event account to path copular sentences, this approach might be successful in accommodating also these facts, but we believe that it is fair to say that the approach has not developed an explicit account of these cases yet.

Another work that follows the same spirit, where estar relates the predication to some external parameters is Sánchez Alonso (2018). Also inspired by Maienborn’s (2005) presuppositional account, Sánchez Alonso proposes that estar, as the marked copulative verb, introduces a boundedness-presupposition. Following Maienborn (2005), ser signals that the circumstance of evaluation i is not the minimal verifying circumstance: in other words, it signals that when the parametric properties of the circumstances of evaluation change – time, space, degree expectation, etc.– the predication with ser continues to be true. In contrast, estar signals that the truth of the evaluation depends on a changeable contextual parameter so that the claim made is boundedly true with respect to one of the contextually determined parameters (Sánchez Alonso 2018: 92): that is, for instance, how she captures that Juan está guapo ‘Juan is^estar handsome’ does not allow us to guarantee that Juan is handsome if the contextual temporal value changes. This boundedness to a contextual parameter can be temporal (referring to a time period, (66a)), spatial (66b) or signal-lowered contextual standards (66c) where a speaker had a previous expectation that is not confirmed (Sánchez Alonso 2018: 95). For instance, in a sentence such as Los zapatos me están anchos ‘The shoes are^estar wide for me’, Sánchez Alonso (2018: 104) argues that estar is possible because context provides a certain degree of wideness that is exceeded by the shoes.

These results are very solid when adjectival attributes are considered (see, in fact, Sánchez Alonso et al. 2017, for experimental results), but they are difficult to extend to the location of events and path structures. The authors would need to argue that events lack any type of contextually-relevant parameters that could vary across time, space or degree, but (see below, Section 4.4) there are clear cases where events involve change across degree values. In this case, for us, the problem is that it becomes very difficult to interpret in which sense a sentence like (65), which is normally taken to be an IL location, involves some sort of contextually determined parameter that makes the sentence boundedly true.

Sánchez Alonso (2018: 96) needs to say that in such cases there is a neutral reading of the verb estar, which follows from the absence of a difference between the expected (spatial) value p and its value in the context of evaluation. It is unclear to us, however, why Spanish would not simply use ser in such cases where using the stronger estar has no effect. Sánchez Alonso’s (2018: 112) approach is extended for locative cases so that what is contextually bound in sentences like (65) is the world parameter, that is, the speaker compares to all other worlds which are identical in any property but the location of the subject, which would take us to the metaphysical question of whether the speaker who says (65) automatically entertains the possibility that Spain might have been on any other continent. We do not discard that such examples can be included in a further extension of the approach, but it is fair to conclude that this integration has not been performed yet.

Finally, Sánchez Alonso (2018) addresses the case of path copular sentences, which she treats as a subcase of the location of events – a position that we have independently adopted in this article. The so-called path copular sentences are, in fact, sentences that locate an entity where an event will take place, for instance, an exiting event in (66).

In her proposal, what is crucial in (66) is that the exit is not at-issue content (Simons et al. 2017), because the content of la salida is presupposed to be known on the common ground. This means that there cannot be falsifying circumstances: the addressee is assumed to know the existence of the exit, and the question is not about the location of that exit, but about how one can get out of the building (Sánchez Alonso 2018: 116). As there can be several ways of exiting the building, one cannot assume that the answer would be the only one that maximally satisfies the circumstances of evaluation, and as such estar cannot be used in this context. This is, however, more complex when the verb appears outside of a question, and a speaker answers something like (67), where it is explicitly stated that there is only one way to exit the building. Ser can still be used in this context.

This is an interesting approach that gives an answer to many of the puzzles that other theories cannot cover, and at least makes an explicit proposal about how it can be extended to the uses of ser with events and paths, which are assumed also to connect to events. However, this is done at the cost of making some potentially problematic assumptions about the intention of the speaker when stating (65) or (66): one has to assume in the first case that the speaker always entertains the possibility that an object could be located in some other place, and in the second case one has to assume that the subject is never at-issue content and other alternatives to perform the event are present.

4.3 Unifying analysis with non-default ser (1): IL predicates, events and temporal intervals

The previous approaches discussed share with many other analyses of the copula the assumption that ser is the default element, and estar is the only one that specifies its selectional restrictions. We start here with an overview of theories that treat ser as equally specified as estar: the two copulas are associated with a contrasting pair of elements, time intervals or degree objects. These approaches avoid overgenerating structures, as they block that ser combines with arguments that in principle must combine with estar. In contrast, they face the problem of how they can group together IL predicates and events to the exclusion of SL predicates, which is empirically required by the distribution of the copulas. Here, we will divide these theories into three main approaches: one where the distinction is based on the nature of the temporal objects that each copula selects (Section 4.3), one where the distinction has to do with the class of comparison or the scale associated with the adjective (Section 4.4), and one based in the more standard IL versus SL distinction (Section 4.5), but attempting to integrate it with locative copular sentences.

To the best of our knowledge, the best example of the first type of analysis is García-Pardo and Menon (2020). These authors propose that the contrast between ser and estar has as its origin the nature of the time arguments of the predicates that are selected by each one of the copulas. The specification of each copula is equally definite: ser combines with predicates whose time argument denotes an interval, irrespective of whether that interval corresponds to an event or a state. In contrast, estar is selected when the time argument associated with the predicate denotes a point in time, that is, a nonextended temporal object (cf. Piñón 1997 for a substantiation of the ontological assumptions behind this, with boundaries and bodies that combine to produce full-fledged temporal objects).

Even if both ser and estar have equally definite selectional restrictions, in line with other authors Fábregas (2012), Zagona (2012), and García-Pardo and Menon (2020) propose that SL predicates are derived from IL predicates. As illustrated in (68), stage-level predicates contain in their structure individual-level predicates. Individual-level predicates consist of a simple predicational structure with only one temporal argument (IL-T). The stage level predicate builds, on top of the predication, a relational PP structure that carries a prepositional head with a WITHIN meaning – roughly comparable to Brucart’s central coincidence relation – which introduces a second temporal argument. The SL-T argument is a point in time (Stage-Level Time) which is located by the P WITHIN as being part of the interval denoted by the IL-T (Individual-Level Time) introduced by Pred.

(68)

IL predicates only have a predicational structure marked with an IL-T, which denotes the whole interval in which the individual manifests a given property (69). Importantly, both SL and IL predicates are different from eventive predicates, which only have an EV-T argument (Event Time) that denotes the whole interval in which the event holds – that is, like IL predicates, the temporal argument of events denotes an interval in (70).

(69)

(70)

As can be seen in (69)–(70), in contrast to (68) the copula emerges as ser. The reason why, according to these authors, is that ser picks time intervals – and it does not care if these time intervals are stative (69) or eventive, as in (70) – whereas estar picks moments in time, i.e. instants (68). This unifying approach splits the domain of temporal objects into two parts, equally positively defined. Its main problem is, however, how to characterize in a unified way the type of temporal object that estar selects. There is no clear evidence that estar chooses instants; prima facie, it is rather the opposite. Note, to begin with, that most estar predicates have a duration that in fact can be explicitly marked with a phrase involving desde ‘since’ or durante ‘for’, which denotes that the situation described by the predicate holds at any single temporal point in the interval defined by the modifier. This is the case in adjectival copulative sentences (71a), locative sentences (71b), and passives (71c).

At the same time, it is not clear that ser always picks intervals; according to the standard characterization of achievements since Vendler (1967), achievements express events that lack a temporal extension. At least initially, one should accordingly expect that event nominals corresponding to achievements (72a) or the eventive passive of achievements (72b) should combine with estar, not with ser, but that expectation goes against the attested facts.

Yet, perhaps the strongest problem of this approach is that, as we have repeatedly observed, it is not true that any locative copular sentence with estar is an SL predicate. Sentences like (73) seem to behave as IL predicates, and in fact intuitively express situations that denote intervals and not points included within an internal, but still take estar. Like Brucart’s (2012) problem that he has to consider any location an SL location, this approach would have to treat any location (not involving an event) as an SL location, against facts such as (73b).

4.4 Unifying analysis with non-default ser (2): scales and comparison classes determine the distribution

In a series of works, Gumiel-Molina and Pérez-Jiménez (2012) and Gumiel-Molina et al. (2015, 2020: Ch. 5) advocate for the idea that the degree properties of the predicate underlie the distribution of the two copulas. These authors have modified their proposal across time: in their (2012) work, inspired by Husband (2010), their proposal was that adjectives whose internal scales are open act as homogeneous properties, while those that have closed scales behave as quantized predicates. Estar would signal the presence of a quantized predicate, that is, a predicate with a closed scale. In (2015) and (2020: Ch. 5), however, their claim – still concentrating on the degree structure of the adjective – is that what is crucial is not the scalar structure of the adjective per se, but the distinction between relative and absolute adjectives reflected on the type of class of comparison that their DegP head introduces as a second argument. The reason is that they find that some adjectives with a closed scale (74) (cf. the modifier completamente ‘completely’, restricted to closed scales; Kennedy and McNally 2005) can combine with ser (Gumiel-Molina et al. 2015: 959).

On the reasonable assumption that the scalar properties of an adjective are lexically determined, they modify the analysis so that the distinction is defined on the functional structure that dominates the adjective, specifically depending on a degree head pos defined as (75) (Gumiel-Molina et al. 2015: 982).

The class of comparison defines a set of properties (P) and, following Fults (2006) it is interpreted as the second argument of the M function that sets the standard degree. Gumiel-Molina et al. (2015) assume that relative adjectives involve comparison between individuals and absolute adjectives involve comparison within an individual, involving different stages of the same subject. This is a distinction that goes back to Crespo (1946) and has Falk (1979) and his distinction between general and individual norms as one of its modern proponents. Given this, the distinction between ser and estar is one that depends on whether the comparison class defines a set of individuals for comparison (between individuals) or different stages or slices of one individual (within one individual) (cf. also Sassoon and Toledo 2011).

We find two problems with this particular approach. In order to account for the use of ser with event subjects (as (12a) above), it is crucial for these authors that events lack scalar structure that allows them to participate in within-individual comparison. The problem is that, according to many theories, scalar structure and degree structure underlie at least some event types, particularly degree achievements (Filip 2008; Hay et al. 1999; Kearns 2007) and predicates with incremental objects that measure the degree of change (Krifka 1986; Leisi 1953; Mourelatos 1978; Tenny 1987; Verkuyl 1972). Specifically, it is possible to construct event nominals built on verbs that contain an incremental change that affects an internal argument, and which is measured according to the degree properties of a base adjective (76). That argument, as far as we can tell, is associated with a within-individual comparison that compares different stages of the individual as the event proceeds.

The equivalent event nominalization, used as a subject, uses ser even though a within-individual comparison is present in the associated event.

Gumiel-Molina et al. (2015) could argue that the degree information is not local to the V node where the copulative is introduced, but note that in their analysis (Gumiel-Molina et al. 2015: 991) the relation between Deg and V does not need to be local: at a minimum, PredP is interposed between the two. In fact, in an example like (77), Deg is closer to V than in a normal copulative structure, because Deg is contained within the subject of predication, which c-commands Pred.

The second problem is that within-individual comparison is impossible in some cases where estar is used. Bazaco (2017) is another example of an analysis that uses degree to explain the distribution of the two copulas. In his case, scalar structure – not comparison classes – is the entity directly associated with the two copulas: estar associates with cales that have an endpoint When an adjective naturally lacks it, a natural process must be inferred that provides a scalar interval to interpret the situation. Admittedly, Bazaco explains that his approach only considers adjectival attributes with degree. One such example of a scalar interval is the so-called evidential use of estar (Mangialavori 2013):

In (78), there is no possible within-individual comparison and there is no access to any slices of the individual. Bazaco’s claim is that what licenses estar here is that one reports his or her experience in interacting with the subject, and that experience defines that the property genial is attained in a higher degree than expected: the evidential use introduces a standout interval of comparison for the subject (Bazaco 2017: 165). Given that there is some distance that is exceeded, an endpoint becomes a relevant part of the semantic interpretation of the scale even for adjectives that lack a closed scale.

Note, however, that, as Bazaco (2017: 92) admits, his scalar-based analysis is restricted to adjectival predicates. Paths are conceptualized often as spatial instances of scales, i.e., sets of exhaustively ordered points across a dimension. In Bazaco’s analysis, the path copular sentences (remember examples like 43) should use estar: the meaning of (79) is that the addressee should follow a path whose endpoint is the location of the subject and as such that path has an endpoint. However, it is constructed with ser.

4.5 Unifying analyses with non-default ser (3): locating events produces an IL predicate

There is a third significant attempt to unify the copulas in a system where none of the copulas is a default element in the strict sense that one of the two verbs lacks any positive specification. Within this view, the location in space or time of an event results in an IL predicate, in the relevant sense that the time and place where an event takes place are in fact inherent and definitional characteristics of an event. Thus, these properties cannot be treated as stage-level properties, that is, as external properties of the situation where the event is placed. To the best of our knowledge, this reasoning has been proposed for Spanish by Leonetti (1994); see also Huyghe and Azzopardi (2019) for a modern manifestation of this idea.

The crucial intuition is that events are intrinsically coded with space and time information. An instantiation of an event is judged as the same or different from another instantiation of the same event according to whether their temporal and spatial coordinates are the same or not. If John and Mary argue at Peter’s party on April the 4th and then again John and Mary argue on their car when going back home two hours later, we interpret that as two different events of John and Mary arguing, while we consider that we are talking of the same John and Mary, even if they are located in different places and times in the two eventualities. From this perspective, the time and the place where an event take place are definitional properties of the event, and any location of an event in space or time would build an IL predicate. Then, (80a) is an instance of an IL predicate, like (80b), and the differences – for instance that (80a) can be quantified over time and (80b) cannot – could follow simply from the nature of the subject involved, which contains event information that allows the quantification; the copula and the predicate would not introduce a temporal variable in either case.

Note that, at any instance, this approach would put the focus of the explanation not on the type of objects with which each one of the copulas combines, but rather on the interpretative result that is obtained once the copula is combined with the predicate and its subject. The approach in principle does not consider that in one case there is eventive information and in the other, there is not. This is so because in this approach a PP like en el tercer piso ‘in the third floor’ would not be grammatical or ungrammatical with either copula until the subject is identified in the structure, and it is the combination of the subject and the predicate – that is, the predication structure as a whole – that would be licensed as IL or SL, and therefore compatible with one or the other copula. To put it differently, in this approach there is no property of the copula or of its complement that has to be licensed by the other element. In this sense, it is only relevant that the property defined in the predicate, applied to that specific subject, counts as IL or SL. Romero (2009), who also emphasizes the role of the subject in the choice of the copula, arrives at a similar conclusion and can be characterized as roughly compatible with the general philosophy of these approaches.

There are, however, some shortcomings of this approach. Starting with a theoretical one, while there is a general consensus that the running time and temporal location of an event can be viewed as inherent properties of the event, it is not so clear that the same cannot be said about the spatial location of some individuals. It is clear that the location of an individual does not define such an individual, but with geographical names (remember for instance (73) above) one could expect that these nouns are defined inherently as having a location, in the same way that events inherently can be argued to have a location.

There might be some way out by introducing a second dimension beyond the IL versus SL property. Arche (2006), who does not address the combination of events with ser, proposes that ser has a classificatory function: ser is used whenever the predicate expresses a class where the individual falls. She proposes that locations are an external property of any individual, irrespective of its type, and this is how she accounts for the need for estar when locating all kinds of individuals. For Arche (2006), since the location of an individual does not define a class of the individual, Juan es en Brasil ‘Juan is^ser in Brazil’ is ungrammatical simply by this rule. Estar would be used, then, for properties that are external to the definitional properties of the individual. Perhaps, if the temporal and spatial location of an event can be interpreted as properties that define classes of the event – or its instantiation s– the approach could be adapted in a way that the distinction between IL and SL in estar locative sentences is reinterpreted as irrelevant.

4.6 Non-unifying analyses: locative copulas as structurally different

One last option would be not to unify the locative copular structures with the adjectival copulative cases. To be clear, the claim in such a type of analysis would be that the underlying structure of the locative copular predicates is substantially different from the one that is obtained in the cases where the copula combines with an adjective. This claim, although not directly made anywhere to the best of our knowledge in the literature on Spanish, is implicit, we believe, in some works that emphasize the existence of different structural conditions. We have already mentioned in passing that Zagona (2012), when acknowledging that her approach overgenerates the set of structures that can appear with ser, suggests that perhaps locative copular structures and ‘normal’ predicative structures with copulative verbs might have different structures. Although she does not provide structures to illustrate exactly what she has in mind or expands on her claim, she suggests that properties of DPs are ‘simple predicates’ while locations are transitive predicates that express a relation between a theme and a DP of location –roughly, a figure and a ground–. If so (and note that this approach would have to reject any type of relational structure for predicates, contra Brucart), one could propose that estar is the default copula for locative predicates while ser is the default copula for simple predicates.

Camacho (2012) also suggests that locative copular sentences have a different structure than other copular structures, proposing that locative copular structures might involve in fact a configuration where the locative is adjoined and not a complement (2012: 471).

Either way, we take it that if the proposal has to involve the claim that the locative copular structure is significantly different from the rest of the uses of the copula, the unification between the uses of the copula has not been reached. At best, in such cases, one would be talking about two structural configurations involving two different types of copulas, with a possible relation of syncretism on the surface that makes the two types of copulas be spelled out by the same set of verbs, ser and estar –unlike what happens, for instance, in languages like Portuguese, where the locative copula and the predicative copula can be expressed by different verbs.

In her extensive typological study, Pustet (2003: 47–54) documents a good number of languages where different copulas are used to express location in time or space, typically in combination with adverbials and adpositional phrases. Spanish, on the surface, does not seem to be this type of language, but if the analysis turns out to confirm that the structure of locative and other types of copular structures is different, the conclusion that Spanish in fact differentiates a set of locative copulas could be granted.

From this non-unificational perspective, the results presented in Section 2 would show that there is an asymmetry between the locative copulas and the other sets: in the locative copulas (81), the copula estar would be the default one, and ser would be the one that has the positive selectional restrictions that focus on events.

In the set of adjectival copulas (82), ser would be the default copula. The two sets of copulas would hold a syncretism relation such as that both would be expressed with the same two verbs, ser and estar, and where presumably the association of ser to the event-taking copula in the locative domain might be facilitated by the general use of ser in predicative contexts as the opposite of a verb estar, that specializes in SL predicates.

As this study does not attempt to provide a unified analysis, its predictions are weaker and therefore it does not face the shortcomings of the other approaches. Its exploration, presumably, would require a broader typological study that focuses on whether Spanish has the syntactic and semantic properties that one finds commonly associated with languages that have locative copulas. With this, we finish our overview of the current proposals.